Home
Search
Collections
Journals
About
Contact us
My IOPscience
Effects of spot size and spot spacing on lateral penumbra reduction when using a dynamic collimation system for spot scanning proton therapy
This content has been downloaded from IOPscience. Please scroll down to see the full text. 2014 Phys. Med. Biol. 59 N187 (http://iopscience.iop.org/0031-9155/59/22/N187) View the table of contents for this issue, or go to the journal homepage for more
Download details: IP Address: 137.149.200.5 This content was downloaded on 11/11/2014 at 22:21
Please note that terms and conditions apply.
Institute of Physics and Engineering in Medicine Phys. Med. Biol. 59 (2014) N187–N196
Physics in Medicine & Biology doi:10.1088/0031-9155/59/22/N187
Note
Effects of spot size and spot spacing on lateral penumbra reduction when using a dynamic collimation system for spot scanning proton therapy Daniel E Hyer, Patrick M Hill, Dongxu Wang, Blake R Smith and Ryan T Flynn Department of Radiation Oncology, University of Iowa, 200 Hawkins Drive, Iowa City, Iowa 52242, USA E-mail: [email protected] Received 23 March 2014, revised 12 August 2014 Accepted for publication 29 August 2014 Published 21 October 2014 Abstract
The purpose of this work was to investigate the reduction in lateral dose penumbra that can be achieved when using a dynamic collimation system (DCS) for spot scanning proton therapy as a function of two beam parameters: spot size and spot spacing. This is an important investigation as both values impact the achievable dose distribution and a wide range of values currently exist depending on delivery hardware. Treatment plans were created both with and without the DCS for in-air spot sizes (σair) of 3, 5, 7, and 9 mm as well as spot spacing intervals of 2, 4, 6 and 8 mm. Compared to un-collimated treatment plans, the plans created with the DCS yielded a reduction in the mean dose to normal tissue surrounding the target of 26.2–40.6% for spot sizes of 3–9 mm, respectively. Increasing the spot spacing resulted in a decrease in the time penalty associated with using the DCS that was approximately proportional to the reduction in the number of rows in the raster delivery pattern. We conclude that dose distributions achievable when using the DCS are comparable to those only attainable with much smaller initial spot sizes, suggesting that the goal of improving high dose conformity may be achieved by either utilizing a DCS or by improving beam line optics. Keywords: spot size, spot spacing, proton, penumbra, collimation, pencil beam scanning (Some figures may appear in colour only in the online journal) S Online supplementary data available from stacks.iop.org/ PMB/59/22N187/mmedia 0031-9155/14/22N187+10$33.00 © 2014 Institute of Physics and Engineering in Medicine Printed in the UK & the USA
N187
D E Hyer et al
Phys. Med. Biol. 59 (2014) N187
1. Introduction In spot scanning (SS) proton therapy, magnetically scanned beams are directed to the appropriate position in the patient and the dose delivered to each resulting beam spot may be individually controlled (Lomax et al 2004). Given the precise spot positioning control in SS, a collimating aperture is typically not used. As a result, the lateral falloff of the dose distribution at the depth of treatment is determined by a combination of the size of the proton beam in air (typically characterized by its spatial standard deviation, σair) and the spread of the beam due to multiple coulomb scattering in the patient. For treatments at shallow depths that require relatively low energy beams, such as in the brain or head and neck, the initial beam size is the main contributing factor to the lateral falloff (Bues et al 2005). Therefore, it is desirable to have a small σair when treating shallow targets in order to limit the dose to normal tissue lateral to the target. The dynamic collimation system (DCS), previously described by Hyer et al (2014), is a new technology that can improve the lateral dose falloff for low energy SS proton therapy. The DCS takes advantage of the fact that a collimation system which shields the entire field-of-view at all points in time (such as an aperture or multi-leaf collimator (MLC)) is not required for SS proton therapy. Instead, the lateral dose distribution can be defined by moving trimmer blades which track the edge of the target and trim the pencil beam as it is scanned in a raster pattern. The main components of the DCS are shown in figure 1 and include two pairs of parallel nickel trimmer blades rotated 90° relative to each other, forming a rectangular shape. Each trimmer blade is driven independently by a linear motor that is connected to the blade via a connecting rod. The trimmer blades are capable of high velocity and acceleration in the direction perpendicular to the central beam axis, 2.5 m s-1 and 19.6 m s-2, respectively. In the current treatment approach, the trimmers remain stationary when the beam is on and the beam is paused when it reaches the center of each delivery row, allowing the blades to be repositioned before the beam continues its scanning pattern. Because the beam is held in a waiting state during the time when the trimmers are moving, there is an increase in treatment delivery time when using the DCS. The increase in treatment delivery time per energy layer is correlated to the number of trimmer movements required, which is proportional to the number of rows and thus dependent on the spot spacing. For this reason, it is important to investigate the effects of spot spacing on treatment delivery time when using a DCS. The second variable investigated in this work was spot size, which is the main contributing factor to the lateral dose falloff at low energies without collimation. This study demonstrates that dose distributions can be improved for a wide range of spot sizes with the use of a DCS, which is important because spot sizes achievable at low energies vary widely among proton therapy systems. For example, the Hitachi system at MD Anderson Cancer Center Proton Therapy Center—Houston has a σair of 14 mm at 72.5 MeV (Gillin et al 2010), the IBA system at the University of Pennsylvania has a σair of 6 mm at 100 MeV (Lin et al 2013), and the custom system at the Paul Scherrer Institute has a σair of 3.5 mm over a wide range of energies (Lomax et al 2004). Another recent publication from MD Anderson has shown that the use of a pre-absorber near the patient surface can reduce the σair value to approximately 9 mm at low energies (Titt et al 2010). To investigate this parameter, SS treatment plans were created and evaluated, both with and without the use of a DCS, over a range of σair values from 3–9 mm. The DCS is still in development and a working prototype system has not yet been constructed. Results presented in this work are based on simulations unless otherwise noted. N188
D E Hyer et al
Phys. Med. Biol. 59 (2014) N187
Figure 1. Major DCS components. The position of each trimmer is individually controlled by a linear motor that is connected to the trimmer via a connecting rod. A range shifter is also mounted just upstream of the trimmers. A support frame and housing surrounding the components is not shown.
2. Methods 2.1. Trimmer positioning
In the current DCS delivery approach, the trimmer blades are positioned such that they block the beam, which is scanning in a rectangular grid pattern, as it arrives at the edge of the target. The exact position of each trimmer blade is chosen through an optimization process that maximizes the ratio of target dose to normal tissue dose. This delivery method may be considered ‘step-and-shoot’ in that the trimmer blades and beam are not in motion simultaneously. Instead, initial trimmer configurations are chosen for the first half row, and the beam is then held in an off state once it arrives at the center of the row until the trimmers are driven to their next position with the linear motors. The beam then continues scanning, with the trimmers remaining stationary, until it arrives at the center of the next row. This delivery technique is qualitatively illustrated in figure 2. The process is repeated until all spots in the energy layer being treated are delivered. A video file illustrating the delivery approach for a single layer is included in the online supplementary files (stacks.iop.org/PMB/59/22N187/mmedia). A full description of the delivery technique has been presented previously (Hyer et al 2014). 2.2. Trimmed pencil beam library
During a treatment, the DCS trimmers will take a wide range of positions relative to the center of the beam, dictated primarily by the shape of the target on the edge being trimmed. In order to generate treatment plans, the optimizer must have access to the beamlet distributions representative of a range of possible DCS configurations. One such method of providing access to these beamlets is to pre-calculate and store the beamlets in a trimmed pencil beam (TPB) library which represents all potential configurations. For this study, TPB libraries were generated using the MCNPX v2.7 Monte Carlo particle transport code for four beam spot sizes and a range of trimmer configurations. N189
D E Hyer et al
Phys. Med. Biol. 59 (2014) N187
Figure 2. (a) A beam’s eye view diagram illustrating the step-and-shoot delivery tech-
nique for positioning the trimmer blades to intercept the beam when it arrives at the edges of the target and (b) selected spot dose distributions qualitatively illustrating the effect of the trimmers on a single spot.
The source was defined by a symmetric non-divergent two-dimensional (2D) Gaussian with the spatial distribution σair of 3, 5, 7, and 9 mm, specified at the entrance of the range shifter. Each TPB library included all possible configurations of X and Y trimmer blade positions starting with edges coincident with the beam central axis (0 mm), retracted in 2 mm increments up to 36 mm, and also positioned at 100 mm (fully retracted, or untrimmed) from the central axis. Thus, each complete TPB library contains the information from a total of 202 =400 simulations. Individual pencil beam dose distributions were generated with the geometry described in figure 3. For each simulation, the proton source was defined with a mean energy E = 125 MeV and Gaussian energy spread of σE = 1 MeV. Energy deposition at the depth of the Bragg peak was recorded using a planar mesh tally 0.5 mm thick and centered at a depth of 3.875 cm. Each simulation tracked 107 source particle histories, which resulted in dose distributions with uncertainties ≤5% for any point receiving ≥10% of the maximum dose. 2.3. Experimental validation of TPBs
To observe the shape that might be expected from TPBs, measurements were performed using the flat edge of a brass aperture to approximate a trimmer blade. A 125 MeV proton beam was passed through a 7.5 g cm−2 acrylic range shifter, past the collimating edge, and into a solid water phantom. The air gap from the range shifter to the phantom surface was set at 25 cm to broaden the spot and accentuate the effects of the collimator edge. The distance from the N190
D E Hyer et al
Phys. Med. Biol. 59 (2014) N187
Figure 3. Geometry of MCNPX simulations used to generate the TPB library.
collimator to the phantom surface was 5 cm. An IBA Lynx detector was placed at the depth of the Bragg peak to measure 2D beam profiles for several beam spots, each with the aperture edge at different distances from the pencil beam central axis. After measurements were completed, Monte Carlo simulations were performed that mimicked the experimental setup to verify the ability of the Monte Carlo model to represent the effects of trimming. 2.4. Generation of treatment plans
2D treatment plans were generated using the TPB library. The dose distributions were limited to 2D since the DCS will predominately affect the dose distribution lateral to the target in any given energy layer. After placing spots in the target area, a least square optimization of spot weights was performed to create dynamically trimmed SS (DTSS) and untrimmed dose distributions for a C-shaped target using σair values of 3, 5, 7, and 9 mm. Equivalent optimization parameters were used to generate DTSS and untrimmed SS dose distributions for comparison purposes. Optimization parameters included mandating that ≥95% of the target area receives 100% of the prescription dose, 110% of the prescription dose, and 100% of the prescription dose. All pixels outside of the target were considered to be normal tissue. A prescription of 100 units of dose and a spot spacing of 2 mm were used to compare both DTSS and untrimmed treatment plans. Additionally, DTSS treatment plans with a spot spacing of 4, 6, and 8 mm were also generated for a σair value of 9 mm. Note that un-collimated SS is normally performed with spot spacing values near σair, but a range of spot spacing values smaller than σair were investigated in this work. This is necessary because the effective σair of the beam spot is greatly reduced with the use of the DCS, as shown in figure 2. N191
D E Hyer et al
Phys. Med. Biol. 59 (2014) N187
Figure 4. Experimental measurements of an untrimmed and half-trimmed profile to
validate the Monte Carlo model.
3. Results 3.1. Experimental results
Figure 4 presents two comparisons between the measured Lynx data and simulated Monte Carlo data in the experimental geometry, one for a full (untrimmed) pencil beam and another for a beam trimmed in half along its central axis. The half-trimmed profiles clearly demonstrate the significant reduction in penumbra made possible by trimming a pencil beam. Overall, the measured data validates the ability of Monte Carlo simulations to characterize the changes in dose spot shape as a result of a trimming device. 3.2. Spot size
Untrimmed SS and DTSS treatment plans for σair values of 3, 5, 7, and 9 mm are shown in figure 5. The dose distributions were calculated in a plane perpendicular to the beam at the depth of the Bragg peak and target coverage was equivalent for the DTSS and untrimmed plans. Figure 6 combines the normal tissue dose-area histograms (DAHs) for all beam sizes, and also shows the mean normal tissue dose as a function of σair for both untrimmed and DTSS plans. The reductions in the mean dose to normal tissue with the use of the DCS were 26.2%, 31.6%, 36.9%, and 40.6% for spot sizes of 3 mm, 5 mm, 7 mm, and 9 mm, respectively. 3.3. Spot spacing
Treatment plans with spot spacing intervals of 2, 4, 6 and 8 mm were generated for a σair value of 9 mm. The results are shown in figure 7. As spot spacing approaches the σair value, it is increasingly harder to achieve uniform target dose coverage in the DTSS plans, and the normal tissue dose also increases. To demonstrate this trend, the normal tissue and target DAHs are plotted in figure 8 for the spot spacing values of 2 and 8 mm. The total estimated time to position the trimmer blades for each of the treatment plans was 4.07 s, 1.95 s, 1.19 s, and 0.93 s for the 2 mm, 4 mm, 6 mm, and 8 mm spot spacing, respectively. N192
D E Hyer et al
Phys. Med. Biol. 59 (2014) N187
Figure 5. 2D dose distributions at the depth of the Bragg Peak for a 125 MeV proton beam with σair values of 3, 5, 7, and 9 mm and a spot spacing of 2 mm. Target edges are shown in black. First row: untrimmed SS dose distributions. Second row: DTSS dose distributions. Third row: subtraction of the DTSS dose distributions from the untrimmed SS dose distributions. Fourth row: DAH of the dose distributions shown in the first and second rows.
Figure 6. (a) The normal tissue DAH for both DTSS and untrimmed treatment plans as a function of initial beam size and (b) mean dose to normal tissue versus spot size for both DTSS and untrimmed treatment plans.
N193
D E Hyer et al
Phys. Med. Biol. 59 (2014) N187
Figure 7. 2D dose distributions at the depth of the Bragg Peak for a 125 MeV proton beam with a constant σair of 9 mm and spot spacing distances of 2, 4, 6 and 8 mm. Target edges are shown in black. First row: untrimmed SS dose distribution. Second row: DTSS dose distribution. Third row: DAH of the dose distributions shown in the first and second rows.
Figure 8. Close up of the target and normal tissue DAH curves for DTSS plans with an initial in-air sigma of 9 mm and spot spacing values of 2 and 8 mm.
4. Discussion To validate our Monte Carlo model, measurements were performed for a representative trimmer configuration. While the geometry and collimating material used for measurement do not match those proposed for the trimmer system, the experimental verification supports the use of Monte Carlo methods to characterize the dose distributions from trimmed beam spots. The minor discrepancies observed between the measured and simulated data in figure 4 are N194
D E Hyer et al
Phys. Med. Biol. 59 (2014) N187
most likely due to uncertainties in the experimental setup. In particular, it was difficult to accurately measure the position of the brass collimator edge during data acquisition. Another potential source of discrepancy could be the multiple coulomb scattering (MCS) model used in MCNPX, which has been previously reported as one of the weaknesses of MCNPX when used for proton simulations (Stankovskiy et al 2009). The angular spread of the beam was also neglected in the model for simplicity and generality. Overall, these discrepancies do not impact the results of this study, as the Monte Carlo results are clearly able to represent the overall dosimetric effects of a trimmer edge on a beamlet. This has allowed TPB libraries to be produced, and enabled the inclusion of 2D treatment plan evaluation in the design process. The improvements in dose distributions achieved by the DCS for a wide range of beam sizes are demonstrated in figure 5. With the addition of the DCS, the dose to the surrounding normal tissue is reduced for all beam sizes, and the reduction increases as the beam size increases. To better illustrate this effect, the normal tissue DAH and the mean dose as a function of σair are plotted in figure 6. For untrimmed plans, the mean dose to surrounding normal tissue increases by 88.3% as the spot size increases from 3 to 9 mm, while the DTSS distributions increase by only 51.6% over the same range. The DTSS approach not only lowers dose to surrounding normal tissue, but also has a weaker dependence on initial spot size. This suggests that the use of a DCS would allow plans of similar dosimetric quality to be created for a wide range of initial beam sizes, spanning those found in proton therapy systems currently in clinical use at a wide range of institutions. The impact of spot spacing when using the DCS was also investigated. As the spot spacing is increased, the untrimmed treatment plans shown in figure 7 show little change. However, the DTSS treatment plans show a visible change in the dose distribution, most notably around the perimeter of the target as expected. For DTSS plans, a uniform dose distribution cannot be achieved with spot spacing values near the initial σair value as the spot size is reduced after interacting with the trimmer blade. The reduction in effective spot size is apparent in figure 6, where a DTSS dose distribution with a 9 mm spot size is found to be dosimetrically equivalent to an untrimmed dose distribution with a spot size of approximately 4 mm. Even with these effects of dose inhomogeneity, the DTSS still offers an improvement over all untrimmed cases with respect to normal tissue dose. The current trimmer positioning algorithm also affects the dose distribution as a function of spot spacing. In the current implementation, a step-and-shoot delivery strategy requires the trimmers to be repositioned only when the beam is held at the center of each row during the raster delivery. This is advantageous in that the trimmers are always positioned and waiting to intercept the beam when it arrives at the edge of the target, eliminating the need to consider more complicated effects of interplay between trimmer and beam motion present in a dynamic delivery. However, such a delivery strategy requires that a single trimmer position collimate two half-rows of the raster scanned delivery (see figure 2). As a result, increases in spot spacing will decrease the apparent resolution of the trimmed dose distribution as a single trimmer position will be used to collimate two half-rows of beam delivery that are now separated by a larger distance. Conceptually, this can be likened to dose distributions that would be achievable with an MLC having leaf widths equal to twice the spot spacing, such that one MLC leaf affects the lateral extent of two rows of delivery. This effect may be seen in figure 7, where the edges of the trimmed dose distributions appear blockier as the spot spacing is increased. For example, one can consider the ‘effective’ MLC leaf width for the 4 mm spot spacing to be approximately 1.2 cm. The advantage of increasing the spot spacing is an associated reduction in the time required to reposition the trimmers, i.e. the overall efficiency of the treatment delivery. Reduction in trimmer positioning time is approximately proportional to the reduction in the number of rows N195
D E Hyer et al
Phys. Med. Biol. 59 (2014) N187
in the raster delivery pattern required to treat the same target area. For example, as the spot spacing is increased from 2 to 4 mm, the number of rows to treat the same target is cut in half and the time required to position the trimmers is reduced by a factor of approximately two (from 4.07 to 1.95 s). When treating stationary targets such as intracranial or head and neck tumors, it is likely that delivery time is less important than overall plan quality, and that an increase in treatment time on the order of seconds per energy layer would be clinically acceptable. However, targets that exhibit motion that may lead to an interplay effect (Seco et al 2009) in the treatment delivery, the speed of delivery will become an increasingly important parameter. In conclusion, there exists a balance between plan quality and delivery efficiency as a function of spot spacing. Large spot spacing reduces the number of trimmer positions which in effect reduces the treatment time, but at the cost of plan quality; small spot spacing achieves better plan quality, but at the cost of an increased number of trimmer positions and subsequently an increase in treatment time. The optimal spot spacing for a given σair value should be determined for each case based on clinical context. 5. Conclusion The DCS is capable of reducing the lateral penumbra in SS proton therapy for a wide range of proton therapy equipment currently in use. The DTSS dose distributions presented in this work yield the greatest improvements for large spot sizes, but improvements can be realized even for σair values as small as 3 mm. Optimal spot spacing considering the balance of plan quality and delivery efficiency should be decided carefully for each clinical case. With the addition of the DCS, users could potentially retrofit older proton systems with large spot sizes to improve dose distributions beyond what is achievable even with the best untrimmed SS system available today. Acknowledgments The authors would like to thank the Central DuPage Hospital Proton Center for providing beam time and staff support for pencil beam measurements. References Bues M, Newhauser W D, Titt U and Smith A R 2005 Therapeutic step and shoot proton beam spotscanning with a multi-leaf collimator: a Monte Carlo study Radiat. Prot. Dosimetry 115 164–9 Gillin M T et al 2010 Commissioning of the discrete spot scanning proton beam delivery system at the University of Texas M.D. Anderson Cancer Center, Proton Therapy Center, Houston Med. Phys. 37 154–63 Hyer D E, Hill P M, Wang D, Smith B R and Flynn R T 2014 A dynamic collimation system for penumbra reduction in spot scanning proton therapy: proof of concept Med. Phys. 41 091701 Lin L, Ainsley C G and McDonough J E 2013 Experimental characterization of 2D pencil beam scanning proton spot profiles Phys. Med. Biol. 58 6193–204 Lomax A J et al 2004 Treatment planning and verification of proton therapy using spot scanning: initial experiences Med. Phys. 31 3150–7 Seco J, Robertson D, Trofimov A and Paganetti H 2009 Breathing interplay effects during proton beam scanning: simulation and statistical analysis Phys. Med. Biol. 54 N283–94 Stankovskiy A, Kerhoas-Cavata S, Ferrand R, Nauraye C and Demarzi L 2009 Monte Carlo modelling of the treatment line of the Proton Therapy Center in Orsay Phys. Med. Biol. 54 2377–94 Titt U et al 2010 Adjustment of the lateral and longitudinal size of scanned proton beam spots using a pre-absorber to optimize penumbrae and delivery efficiency Phys. Med. Biol. 55 7097–106 N196
Search
Collections
Journals
About
Contact us
My IOPscience
Effects of spot size and spot spacing on lateral penumbra reduction when using a dynamic collimation system for spot scanning proton therapy
This content has been downloaded from IOPscience. Please scroll down to see the full text. 2014 Phys. Med. Biol. 59 N187 (http://iopscience.iop.org/0031-9155/59/22/N187) View the table of contents for this issue, or go to the journal homepage for more
Download details: IP Address: 137.149.200.5 This content was downloaded on 11/11/2014 at 22:21
Please note that terms and conditions apply.
Institute of Physics and Engineering in Medicine Phys. Med. Biol. 59 (2014) N187–N196
Physics in Medicine & Biology doi:10.1088/0031-9155/59/22/N187
Note
Effects of spot size and spot spacing on lateral penumbra reduction when using a dynamic collimation system for spot scanning proton therapy Daniel E Hyer, Patrick M Hill, Dongxu Wang, Blake R Smith and Ryan T Flynn Department of Radiation Oncology, University of Iowa, 200 Hawkins Drive, Iowa City, Iowa 52242, USA E-mail: [email protected] Received 23 March 2014, revised 12 August 2014 Accepted for publication 29 August 2014 Published 21 October 2014 Abstract
The purpose of this work was to investigate the reduction in lateral dose penumbra that can be achieved when using a dynamic collimation system (DCS) for spot scanning proton therapy as a function of two beam parameters: spot size and spot spacing. This is an important investigation as both values impact the achievable dose distribution and a wide range of values currently exist depending on delivery hardware. Treatment plans were created both with and without the DCS for in-air spot sizes (σair) of 3, 5, 7, and 9 mm as well as spot spacing intervals of 2, 4, 6 and 8 mm. Compared to un-collimated treatment plans, the plans created with the DCS yielded a reduction in the mean dose to normal tissue surrounding the target of 26.2–40.6% for spot sizes of 3–9 mm, respectively. Increasing the spot spacing resulted in a decrease in the time penalty associated with using the DCS that was approximately proportional to the reduction in the number of rows in the raster delivery pattern. We conclude that dose distributions achievable when using the DCS are comparable to those only attainable with much smaller initial spot sizes, suggesting that the goal of improving high dose conformity may be achieved by either utilizing a DCS or by improving beam line optics. Keywords: spot size, spot spacing, proton, penumbra, collimation, pencil beam scanning (Some figures may appear in colour only in the online journal) S Online supplementary data available from stacks.iop.org/ PMB/59/22N187/mmedia 0031-9155/14/22N187+10$33.00 © 2014 Institute of Physics and Engineering in Medicine Printed in the UK & the USA
N187
D E Hyer et al
Phys. Med. Biol. 59 (2014) N187
1. Introduction In spot scanning (SS) proton therapy, magnetically scanned beams are directed to the appropriate position in the patient and the dose delivered to each resulting beam spot may be individually controlled (Lomax et al 2004). Given the precise spot positioning control in SS, a collimating aperture is typically not used. As a result, the lateral falloff of the dose distribution at the depth of treatment is determined by a combination of the size of the proton beam in air (typically characterized by its spatial standard deviation, σair) and the spread of the beam due to multiple coulomb scattering in the patient. For treatments at shallow depths that require relatively low energy beams, such as in the brain or head and neck, the initial beam size is the main contributing factor to the lateral falloff (Bues et al 2005). Therefore, it is desirable to have a small σair when treating shallow targets in order to limit the dose to normal tissue lateral to the target. The dynamic collimation system (DCS), previously described by Hyer et al (2014), is a new technology that can improve the lateral dose falloff for low energy SS proton therapy. The DCS takes advantage of the fact that a collimation system which shields the entire field-of-view at all points in time (such as an aperture or multi-leaf collimator (MLC)) is not required for SS proton therapy. Instead, the lateral dose distribution can be defined by moving trimmer blades which track the edge of the target and trim the pencil beam as it is scanned in a raster pattern. The main components of the DCS are shown in figure 1 and include two pairs of parallel nickel trimmer blades rotated 90° relative to each other, forming a rectangular shape. Each trimmer blade is driven independently by a linear motor that is connected to the blade via a connecting rod. The trimmer blades are capable of high velocity and acceleration in the direction perpendicular to the central beam axis, 2.5 m s-1 and 19.6 m s-2, respectively. In the current treatment approach, the trimmers remain stationary when the beam is on and the beam is paused when it reaches the center of each delivery row, allowing the blades to be repositioned before the beam continues its scanning pattern. Because the beam is held in a waiting state during the time when the trimmers are moving, there is an increase in treatment delivery time when using the DCS. The increase in treatment delivery time per energy layer is correlated to the number of trimmer movements required, which is proportional to the number of rows and thus dependent on the spot spacing. For this reason, it is important to investigate the effects of spot spacing on treatment delivery time when using a DCS. The second variable investigated in this work was spot size, which is the main contributing factor to the lateral dose falloff at low energies without collimation. This study demonstrates that dose distributions can be improved for a wide range of spot sizes with the use of a DCS, which is important because spot sizes achievable at low energies vary widely among proton therapy systems. For example, the Hitachi system at MD Anderson Cancer Center Proton Therapy Center—Houston has a σair of 14 mm at 72.5 MeV (Gillin et al 2010), the IBA system at the University of Pennsylvania has a σair of 6 mm at 100 MeV (Lin et al 2013), and the custom system at the Paul Scherrer Institute has a σair of 3.5 mm over a wide range of energies (Lomax et al 2004). Another recent publication from MD Anderson has shown that the use of a pre-absorber near the patient surface can reduce the σair value to approximately 9 mm at low energies (Titt et al 2010). To investigate this parameter, SS treatment plans were created and evaluated, both with and without the use of a DCS, over a range of σair values from 3–9 mm. The DCS is still in development and a working prototype system has not yet been constructed. Results presented in this work are based on simulations unless otherwise noted. N188
D E Hyer et al
Phys. Med. Biol. 59 (2014) N187
Figure 1. Major DCS components. The position of each trimmer is individually controlled by a linear motor that is connected to the trimmer via a connecting rod. A range shifter is also mounted just upstream of the trimmers. A support frame and housing surrounding the components is not shown.
2. Methods 2.1. Trimmer positioning
In the current DCS delivery approach, the trimmer blades are positioned such that they block the beam, which is scanning in a rectangular grid pattern, as it arrives at the edge of the target. The exact position of each trimmer blade is chosen through an optimization process that maximizes the ratio of target dose to normal tissue dose. This delivery method may be considered ‘step-and-shoot’ in that the trimmer blades and beam are not in motion simultaneously. Instead, initial trimmer configurations are chosen for the first half row, and the beam is then held in an off state once it arrives at the center of the row until the trimmers are driven to their next position with the linear motors. The beam then continues scanning, with the trimmers remaining stationary, until it arrives at the center of the next row. This delivery technique is qualitatively illustrated in figure 2. The process is repeated until all spots in the energy layer being treated are delivered. A video file illustrating the delivery approach for a single layer is included in the online supplementary files (stacks.iop.org/PMB/59/22N187/mmedia). A full description of the delivery technique has been presented previously (Hyer et al 2014). 2.2. Trimmed pencil beam library
During a treatment, the DCS trimmers will take a wide range of positions relative to the center of the beam, dictated primarily by the shape of the target on the edge being trimmed. In order to generate treatment plans, the optimizer must have access to the beamlet distributions representative of a range of possible DCS configurations. One such method of providing access to these beamlets is to pre-calculate and store the beamlets in a trimmed pencil beam (TPB) library which represents all potential configurations. For this study, TPB libraries were generated using the MCNPX v2.7 Monte Carlo particle transport code for four beam spot sizes and a range of trimmer configurations. N189
D E Hyer et al
Phys. Med. Biol. 59 (2014) N187
Figure 2. (a) A beam’s eye view diagram illustrating the step-and-shoot delivery tech-
nique for positioning the trimmer blades to intercept the beam when it arrives at the edges of the target and (b) selected spot dose distributions qualitatively illustrating the effect of the trimmers on a single spot.
The source was defined by a symmetric non-divergent two-dimensional (2D) Gaussian with the spatial distribution σair of 3, 5, 7, and 9 mm, specified at the entrance of the range shifter. Each TPB library included all possible configurations of X and Y trimmer blade positions starting with edges coincident with the beam central axis (0 mm), retracted in 2 mm increments up to 36 mm, and also positioned at 100 mm (fully retracted, or untrimmed) from the central axis. Thus, each complete TPB library contains the information from a total of 202 =400 simulations. Individual pencil beam dose distributions were generated with the geometry described in figure 3. For each simulation, the proton source was defined with a mean energy E = 125 MeV and Gaussian energy spread of σE = 1 MeV. Energy deposition at the depth of the Bragg peak was recorded using a planar mesh tally 0.5 mm thick and centered at a depth of 3.875 cm. Each simulation tracked 107 source particle histories, which resulted in dose distributions with uncertainties ≤5% for any point receiving ≥10% of the maximum dose. 2.3. Experimental validation of TPBs
To observe the shape that might be expected from TPBs, measurements were performed using the flat edge of a brass aperture to approximate a trimmer blade. A 125 MeV proton beam was passed through a 7.5 g cm−2 acrylic range shifter, past the collimating edge, and into a solid water phantom. The air gap from the range shifter to the phantom surface was set at 25 cm to broaden the spot and accentuate the effects of the collimator edge. The distance from the N190
D E Hyer et al
Phys. Med. Biol. 59 (2014) N187
Figure 3. Geometry of MCNPX simulations used to generate the TPB library.
collimator to the phantom surface was 5 cm. An IBA Lynx detector was placed at the depth of the Bragg peak to measure 2D beam profiles for several beam spots, each with the aperture edge at different distances from the pencil beam central axis. After measurements were completed, Monte Carlo simulations were performed that mimicked the experimental setup to verify the ability of the Monte Carlo model to represent the effects of trimming. 2.4. Generation of treatment plans
2D treatment plans were generated using the TPB library. The dose distributions were limited to 2D since the DCS will predominately affect the dose distribution lateral to the target in any given energy layer. After placing spots in the target area, a least square optimization of spot weights was performed to create dynamically trimmed SS (DTSS) and untrimmed dose distributions for a C-shaped target using σair values of 3, 5, 7, and 9 mm. Equivalent optimization parameters were used to generate DTSS and untrimmed SS dose distributions for comparison purposes. Optimization parameters included mandating that ≥95% of the target area receives 100% of the prescription dose, 110% of the prescription dose, and 100% of the prescription dose. All pixels outside of the target were considered to be normal tissue. A prescription of 100 units of dose and a spot spacing of 2 mm were used to compare both DTSS and untrimmed treatment plans. Additionally, DTSS treatment plans with a spot spacing of 4, 6, and 8 mm were also generated for a σair value of 9 mm. Note that un-collimated SS is normally performed with spot spacing values near σair, but a range of spot spacing values smaller than σair were investigated in this work. This is necessary because the effective σair of the beam spot is greatly reduced with the use of the DCS, as shown in figure 2. N191
D E Hyer et al
Phys. Med. Biol. 59 (2014) N187
Figure 4. Experimental measurements of an untrimmed and half-trimmed profile to
validate the Monte Carlo model.
3. Results 3.1. Experimental results
Figure 4 presents two comparisons between the measured Lynx data and simulated Monte Carlo data in the experimental geometry, one for a full (untrimmed) pencil beam and another for a beam trimmed in half along its central axis. The half-trimmed profiles clearly demonstrate the significant reduction in penumbra made possible by trimming a pencil beam. Overall, the measured data validates the ability of Monte Carlo simulations to characterize the changes in dose spot shape as a result of a trimming device. 3.2. Spot size
Untrimmed SS and DTSS treatment plans for σair values of 3, 5, 7, and 9 mm are shown in figure 5. The dose distributions were calculated in a plane perpendicular to the beam at the depth of the Bragg peak and target coverage was equivalent for the DTSS and untrimmed plans. Figure 6 combines the normal tissue dose-area histograms (DAHs) for all beam sizes, and also shows the mean normal tissue dose as a function of σair for both untrimmed and DTSS plans. The reductions in the mean dose to normal tissue with the use of the DCS were 26.2%, 31.6%, 36.9%, and 40.6% for spot sizes of 3 mm, 5 mm, 7 mm, and 9 mm, respectively. 3.3. Spot spacing
Treatment plans with spot spacing intervals of 2, 4, 6 and 8 mm were generated for a σair value of 9 mm. The results are shown in figure 7. As spot spacing approaches the σair value, it is increasingly harder to achieve uniform target dose coverage in the DTSS plans, and the normal tissue dose also increases. To demonstrate this trend, the normal tissue and target DAHs are plotted in figure 8 for the spot spacing values of 2 and 8 mm. The total estimated time to position the trimmer blades for each of the treatment plans was 4.07 s, 1.95 s, 1.19 s, and 0.93 s for the 2 mm, 4 mm, 6 mm, and 8 mm spot spacing, respectively. N192
D E Hyer et al
Phys. Med. Biol. 59 (2014) N187
Figure 5. 2D dose distributions at the depth of the Bragg Peak for a 125 MeV proton beam with σair values of 3, 5, 7, and 9 mm and a spot spacing of 2 mm. Target edges are shown in black. First row: untrimmed SS dose distributions. Second row: DTSS dose distributions. Third row: subtraction of the DTSS dose distributions from the untrimmed SS dose distributions. Fourth row: DAH of the dose distributions shown in the first and second rows.
Figure 6. (a) The normal tissue DAH for both DTSS and untrimmed treatment plans as a function of initial beam size and (b) mean dose to normal tissue versus spot size for both DTSS and untrimmed treatment plans.
N193
D E Hyer et al
Phys. Med. Biol. 59 (2014) N187
Figure 7. 2D dose distributions at the depth of the Bragg Peak for a 125 MeV proton beam with a constant σair of 9 mm and spot spacing distances of 2, 4, 6 and 8 mm. Target edges are shown in black. First row: untrimmed SS dose distribution. Second row: DTSS dose distribution. Third row: DAH of the dose distributions shown in the first and second rows.
Figure 8. Close up of the target and normal tissue DAH curves for DTSS plans with an initial in-air sigma of 9 mm and spot spacing values of 2 and 8 mm.
4. Discussion To validate our Monte Carlo model, measurements were performed for a representative trimmer configuration. While the geometry and collimating material used for measurement do not match those proposed for the trimmer system, the experimental verification supports the use of Monte Carlo methods to characterize the dose distributions from trimmed beam spots. The minor discrepancies observed between the measured and simulated data in figure 4 are N194
D E Hyer et al
Phys. Med. Biol. 59 (2014) N187
most likely due to uncertainties in the experimental setup. In particular, it was difficult to accurately measure the position of the brass collimator edge during data acquisition. Another potential source of discrepancy could be the multiple coulomb scattering (MCS) model used in MCNPX, which has been previously reported as one of the weaknesses of MCNPX when used for proton simulations (Stankovskiy et al 2009). The angular spread of the beam was also neglected in the model for simplicity and generality. Overall, these discrepancies do not impact the results of this study, as the Monte Carlo results are clearly able to represent the overall dosimetric effects of a trimmer edge on a beamlet. This has allowed TPB libraries to be produced, and enabled the inclusion of 2D treatment plan evaluation in the design process. The improvements in dose distributions achieved by the DCS for a wide range of beam sizes are demonstrated in figure 5. With the addition of the DCS, the dose to the surrounding normal tissue is reduced for all beam sizes, and the reduction increases as the beam size increases. To better illustrate this effect, the normal tissue DAH and the mean dose as a function of σair are plotted in figure 6. For untrimmed plans, the mean dose to surrounding normal tissue increases by 88.3% as the spot size increases from 3 to 9 mm, while the DTSS distributions increase by only 51.6% over the same range. The DTSS approach not only lowers dose to surrounding normal tissue, but also has a weaker dependence on initial spot size. This suggests that the use of a DCS would allow plans of similar dosimetric quality to be created for a wide range of initial beam sizes, spanning those found in proton therapy systems currently in clinical use at a wide range of institutions. The impact of spot spacing when using the DCS was also investigated. As the spot spacing is increased, the untrimmed treatment plans shown in figure 7 show little change. However, the DTSS treatment plans show a visible change in the dose distribution, most notably around the perimeter of the target as expected. For DTSS plans, a uniform dose distribution cannot be achieved with spot spacing values near the initial σair value as the spot size is reduced after interacting with the trimmer blade. The reduction in effective spot size is apparent in figure 6, where a DTSS dose distribution with a 9 mm spot size is found to be dosimetrically equivalent to an untrimmed dose distribution with a spot size of approximately 4 mm. Even with these effects of dose inhomogeneity, the DTSS still offers an improvement over all untrimmed cases with respect to normal tissue dose. The current trimmer positioning algorithm also affects the dose distribution as a function of spot spacing. In the current implementation, a step-and-shoot delivery strategy requires the trimmers to be repositioned only when the beam is held at the center of each row during the raster delivery. This is advantageous in that the trimmers are always positioned and waiting to intercept the beam when it arrives at the edge of the target, eliminating the need to consider more complicated effects of interplay between trimmer and beam motion present in a dynamic delivery. However, such a delivery strategy requires that a single trimmer position collimate two half-rows of the raster scanned delivery (see figure 2). As a result, increases in spot spacing will decrease the apparent resolution of the trimmed dose distribution as a single trimmer position will be used to collimate two half-rows of beam delivery that are now separated by a larger distance. Conceptually, this can be likened to dose distributions that would be achievable with an MLC having leaf widths equal to twice the spot spacing, such that one MLC leaf affects the lateral extent of two rows of delivery. This effect may be seen in figure 7, where the edges of the trimmed dose distributions appear blockier as the spot spacing is increased. For example, one can consider the ‘effective’ MLC leaf width for the 4 mm spot spacing to be approximately 1.2 cm. The advantage of increasing the spot spacing is an associated reduction in the time required to reposition the trimmers, i.e. the overall efficiency of the treatment delivery. Reduction in trimmer positioning time is approximately proportional to the reduction in the number of rows N195
D E Hyer et al
Phys. Med. Biol. 59 (2014) N187
in the raster delivery pattern required to treat the same target area. For example, as the spot spacing is increased from 2 to 4 mm, the number of rows to treat the same target is cut in half and the time required to position the trimmers is reduced by a factor of approximately two (from 4.07 to 1.95 s). When treating stationary targets such as intracranial or head and neck tumors, it is likely that delivery time is less important than overall plan quality, and that an increase in treatment time on the order of seconds per energy layer would be clinically acceptable. However, targets that exhibit motion that may lead to an interplay effect (Seco et al 2009) in the treatment delivery, the speed of delivery will become an increasingly important parameter. In conclusion, there exists a balance between plan quality and delivery efficiency as a function of spot spacing. Large spot spacing reduces the number of trimmer positions which in effect reduces the treatment time, but at the cost of plan quality; small spot spacing achieves better plan quality, but at the cost of an increased number of trimmer positions and subsequently an increase in treatment time. The optimal spot spacing for a given σair value should be determined for each case based on clinical context. 5. Conclusion The DCS is capable of reducing the lateral penumbra in SS proton therapy for a wide range of proton therapy equipment currently in use. The DTSS dose distributions presented in this work yield the greatest improvements for large spot sizes, but improvements can be realized even for σair values as small as 3 mm. Optimal spot spacing considering the balance of plan quality and delivery efficiency should be decided carefully for each clinical case. With the addition of the DCS, users could potentially retrofit older proton systems with large spot sizes to improve dose distributions beyond what is achievable even with the best untrimmed SS system available today. Acknowledgments The authors would like to thank the Central DuPage Hospital Proton Center for providing beam time and staff support for pencil beam measurements. References Bues M, Newhauser W D, Titt U and Smith A R 2005 Therapeutic step and shoot proton beam spotscanning with a multi-leaf collimator: a Monte Carlo study Radiat. Prot. Dosimetry 115 164–9 Gillin M T et al 2010 Commissioning of the discrete spot scanning proton beam delivery system at the University of Texas M.D. Anderson Cancer Center, Proton Therapy Center, Houston Med. Phys. 37 154–63 Hyer D E, Hill P M, Wang D, Smith B R and Flynn R T 2014 A dynamic collimation system for penumbra reduction in spot scanning proton therapy: proof of concept Med. Phys. 41 091701 Lin L, Ainsley C G and McDonough J E 2013 Experimental characterization of 2D pencil beam scanning proton spot profiles Phys. Med. Biol. 58 6193–204 Lomax A J et al 2004 Treatment planning and verification of proton therapy using spot scanning: initial experiences Med. Phys. 31 3150–7 Seco J, Robertson D, Trofimov A and Paganetti H 2009 Breathing interplay effects during proton beam scanning: simulation and statistical analysis Phys. Med. Biol. 54 N283–94 Stankovskiy A, Kerhoas-Cavata S, Ferrand R, Nauraye C and Demarzi L 2009 Monte Carlo modelling of the treatment line of the Proton Therapy Center in Orsay Phys. Med. Biol. 54 2377–94 Titt U et al 2010 Adjustment of the lateral and longitudinal size of scanned proton beam spots using a pre-absorber to optimize penumbrae and delivery efficiency Phys. Med. Biol. 55 7097–106 N196
